W3: Data Cleaning 2

Agenda:

  • Work Time (20 minutes)
  • Introduce State Cancer Profiles project
  • Data Cleaning 2: Tidy Data

{janitor}

Oftentimes, capitilization and spacing in column names is a nightmare when bringing in data.

  • cutoff 20%
  • FruitPerDay

And it is hard to consistently call things. We need to use backticks when we refer to variables.

What janitor::clean_names() is transform your variable names into consistent capitalization and spacing.

  • cutoff_20
  • fruit_per_day

It’s usually one of the first functions that I use when I load data.

Tidy Data

A standard of organizing data ready for analysis: Tidy data.

Many tools in R, such as plotting, are most effective when your data is in a Tidy format.

What is Tidy Data?

  1. Each observation must have its own row.
  2. Each variable must have its own column.
  3. Each value must have its own cell.
A tidy dataframe.

Easier to understand by looking at common un-Tidy data situations.

Tidy or Not?

  • Depends on what you want to do with the data
  • Plot fertility rates over years, with a line for each country (not tidy)

Situations with untidy data

  • Multiple columns map to a single variable (wide -> long)
  • One column maps to multiple variables (long -> wide)
  • One column contains mutliple values (separate)

The solution for this is the {tidyr} package!

{tidyr}

  • Part of the tidyverse
  • Contains functions for reshaping data
    • pivot_longer()
    • pivot_wider()
    • separate()
    • drop_na()

Situation: Multiple columns map to a single variable

untidy1
  • Want to look at the relationship between sales numbers and quarters for each store type.
  • Need variables: Quarter, Sales numbers, Store type
  • 👌 Observation is a store, and each observation has its own row.
  • 😩 The columns “Q1_Sales”, …, “Q4_Sales” seem to be values of a single variable “quarter” of our observation.

Reshaping the data

  • 😩 The columns “Q1_Sales”, …, “Q4_Sales” seem to be values of a single variable “quarter” of our observation.
  • We want to make these into two columns:
    • The quarter name (Quarter),
    • The value for that quarter (Sales)

pivot_longer()

Solution: make the data into a longer format.

Tidyexplain

pivot_longer()

Solution: make the data into a longer format.

tidy1 = untidy1 |>
  pivot_longer(cols= c("Q1_Sales", 
                       "Q2_Sales", 
                       "Q3_Sales", 
                       "Q4_Sales"), 
               names_to = "Quarter", 
               values_to = "Sales")
tidy1

Tidyexplain

pivot_longer()

tidy1
  • 👌 Observation is a store’s quarter.
  • 👌 The new columns “quarter” and “sales” are variables with their own columns.

Plotting the longer data

This is a common transformation, as it is easier to do data entry via a wider format, but the tools we use in programming often requires it in a longer format.

tidy1
  • Try plotting the transformed data tidy1
    • map x to Quarter, y to Sales, and group to Store
ggplot(tidy1) + aes(x = Quarter, y = Sales, group = Store) + geom_line() 
ggplot(tidy1) + aes(x = Quarter, 
                    y = Sales, 
                    group = Store) + 
  geom_line()

Thinking about pivot_longer()

  • What columns in this dataset could be combined into one column?
relig_income_sm[1:5,]
  • You’ll have the opportunity to work on this at the end of class.

Situation: One column maps to multiple variables

untidy2
  • What’s the relationship between KRAS mutation and KRAS expression?
  • 😩 Observation is a sample’s gene type?
  • 😩 The variables we want, KRAS_mutation and KRAS_expression, are in rows. The current columns contain multiple types of info: gene contains both mutations and expression, and value contains both gene expression and mutational status.

pivot_wider()

Solution: make the data into a wider format. Tidyexplain

.

pivot_wider()

Solution: make the data into a wider format. Tidyexplain

tidy2 = pivot_wider(untidy2, 
                    names_from = "gene", 
                    values_from = "values") |>
  mutate(KRAS_mutation = as.logical(KRAS_mutation))

tidy2

pivot_wider()

tidy2

👌 Observation is a sample. 👌 The new columns are appropriate variables for our analysis.

Plotting our wide data

tidy2
  • Take a look at tidy2 first.
  • Make a ggplot with tidy2.
    • Map x to KRAS_mutation
    • y to KRAS_expression
    • color to KRAS_mutation
ggplot(tidy2) + aes(x = KRAS_mutation, y = KRAS_expression, color = KRAS_mutation) + geom_point() 
ggplot(tidy2) + aes(x = KRAS_mutation, 
                    y = KRAS_expression,
                    color = KRAS_mutation) + 
  geom_point()

Inverse Operations

pivot_longer() can undo what we did in pivot_wider(), and vice versa.

pivot_longer(tidy2, c("KRAS_mutation", 
                      "KRAS_expression"), 
             names_to = "gene", 
             values_to = "values")

Situation: Multiple variables are stored in a single column

untidy3

😩 The rate column’s values have multiple values in its cell.

separate()

Solution: Let’s separate it:

Before

untidy3

After

tidy3 = separate(untidy3, 
                 col = "rate", 
                 into = c("count", "population"), 
                 sep = "/")
tidy3

Subjectivity in Tidy data

We have looked at clear cases of when a dataset is Tidy. In reality, the Tidy state depends on what we call variables and observations for the analysis we want to conduct.

Tools such as ggplot require precise definition of our variables, so planning ahead what we want to use with our tools creates clarity of what we call variables and observations.

Tip: think about what you want to do with the data, and work backwards. That will help you identify whether the data is tidy or not.

Challenge

What analysis would you want to do with this dataset, and what kind of transformation would you do to get it Tidy?

relig_income_sm |> pivot_longer(cols = c(`<$10k`,`$10-20k`,`$20-30k` ), names_to = "income", values_to = "count", ) 
relig_income_sm |>
  pivot_longer(cols = c(`<$10k`,`$10-20k`,`$20-30k` ),
    names_to = "income",
    values_to = "count",
    )